A truly brilliant comparison with good balanced scoring making it very accurate, judgement wise. The end of the video rightly calls out the continuing weakness of Firefly in this area - especially for any professional work. I look forward to more 'nailed it' comparisons. Succint and to the point. Brilliant.
Great comparison, Blake. If only Magnific had a subscription level below Pro ($39/month). There's no way I'd do 200 upscales or generations a month (what's included in the Pro level). Wish they had a Basic level for more like $15-20/month.
I agree! I got it back in Dec 2023 and they gave me 30,000 credits. I am currently at 24,000. Many things require anywhere from 7 - 100 credits per generation depending on the resolution. So it is VERY difficult to hit that pro level even for someone like me who generates a lot of images.
@ Yikes! That's a lot of remaining credits at the end of the year. When it renews in a couple months you'll have about 53,000 credits. 😬 Too bad I can't buy $40 worth of credits and replenish them once I've used them up in 9 months. Wanna sell me some of your extra credits on the black market? Haha.
A very well thought out and well presented look at these things. I know a-lot more now than I did 30 min ago. I'm so disappointed in Adobe with what you showed. And your comments around the 17 min mark are spot on. I use or at least try to use Gen remove in both Lr and PS often. Even after adobe max updates it's quite limited. Thanks - Tim
Enjoyable. So many AI generators to work with. Sometimes the errors or the misunderstanding can be the funniest things. I find that knowing PS and other Post processing software help tremedousely with the final results. Also MJ's new web based editor has some cool features. Thanks for putting this together was a fun and informative watch. :)
I agree. Knowing Ps makes AI image generation a much more enjoyable experience and you can edit them to your ideal vision. I haven't played with the web module for MJ, only the Discord.
Thanks for the comparisons, have you any experience with these using your own photo then using AI to fill in the background with prompts like generative fill does in photoshop? I have noticed if I use prompts to give me an illustrative background it can work magic however for work i need a realistic background. Thanks for any suggestions.
I haven't used it in my work in that way. However, magnific would be best for that. Even if slightly illustrative, blurred a little it may look natural.
For fun I tried same wizard prompt in Midjourney and got pretty decent result - point is that I'm using my specific personalized styles. Without personalization styles images are ... OK but not that good.
Great review Blake! Thanks. I mostly use Midjourney and have found if I say Photo or Photograph rather than Photo Realistic I get better results. Realistic seems to be some kind of trigger that it’s not a real photo but photo like.
You get credits. 30,000 when you become pro. It's 7 credits for 2k resolution images and 22 credits for 4k resolution images. Up sizing can be anywhere from 5 - 75 credits based on how large you upscale it. I got my account December 2023 and still have over 20,000 credits. It's safe to say you can generate a LOT with the Pro account.
That was a typo, my bad, but hey it put the limits of the AI to the test and it still produced wonderful results... Shows that it's smart even when human input is NOT, lol
Just so you know, photorealistic is an art style which makes image closer to realism. If you want a realistic image, dont use photorealistic in the prompt.
@@f64Academy just add realistic or cinematic or cinematic still etc. using words like portrait photography, nature photography etc. in the prompt will also help moving ai towards realistic images. Not every time though sometimes ai more bends towards art styles as per its understanding of the prompt.
Nice work thru AI stuff, Blake. You are totally right. It will only get better, and it is here to stay. I think for compositing it is very usefull. What that can evolve to in 5 years is exiting to think of, it is only a coupple of years sincewe even heard about it.
For some reason I always get a slight variation of the same woman's face every time I use Magnific (Mystic/Flux). And I don't seem to be the only one who has this issue because I see her face everywhere in videos about Flux. I agree that the overall quality surpasses even Midjourney but unless I can figure out how to avoid seeing that same face over and over, I have to stick with Midjourney and then upscale in Magnific.
That's still the best plan, I think. there's more versatility in the prompts from MJ. So picking the one you want is a good option, then upscaling it with Magnific as MJ's upscale isn't the best.
@@f64Academy Now if Photoshop could just get a Midjourney or Flux plugin that bypasses Firefly, we'd be all set. I agree with everything you said about Firefly. Adobe needs to rethink things.
@@f64Academy One last thing - I use all of these tools as well - and even a few others - on a near, daily basis. The one dark horse no ones talking about is "Visual Electric". I've been using it for about a Month now - and in MANY cases - it outperforms Midjourney or at the very least adheres to the prompt way better. At certain creative and/or cinematic images MJ still outshines the rest, but VE is my go to for people and my own stock-type images. Also, the tech help is super responsive and i LOVE the whole "whiteboard canvas" setup they have. Let's you explore lots of options for a given client all in one "page". (and NO, i don't work for them in any capacity and i don't get anything if you go check it out!)
Great review but Adobe may getting a bad rap. I just downloaded Photoshop (beta) 25.1.0 and entered your prompts for each example in Generative Workspace. All four options from Photoshop's Firefly Image 3 were as good or better than the competitor's except for the "Nailed It" logo. Photoshop did a great job on the hammer and nail but messed up royally on the words "Nailed It" in each option.
Maybe in the direction of accuracy to prompt, but the details are always horrible and always will be until they release the 1024 restriction. It's not useable. I have the exact same version of PS installed and those were the exact images I received for my prompts and I made the canvas size 1024x768... All horrible.
Well this was interesting. I am blown away with how bad Adobe Firefly is. I use the generative fill to mostly do removal related tasks and maybe adding a plant here and there. Works okay for that but not great. With all the money adobe has coming in through subscriptions and all the photos they have access to via there not so user friendly EULA, and all the stock images they have access to on top of that, it is really pathetic what they have to offer. Looking forward to working with these AI tools over the winter.
Yes, it's true, these algorithms accidentally named AI will remain with us and will be increasingly perfect in creating imaginary worlds. However, nothing can replace a human in creativity, which is the food for these algorithms and there is no point in complaining because we are the ones who feed these algorithms with our creativity. Our activity has an impact on the quality of the images created. Complaining about AI, we complain about our own creativity, which people forget about. I do not use such options, but I follow its progress. Greetings from Poland from the city of "Przemyśl Fortress".
Just so you know, Magnific uses the Flux model, which is available in dozens of places. It is not exclusive.
It's an awesome model. Thanks for the Intel.
How did you find out about the Flux model? Where else is it used? Do you have any pointers for me where I could start my research? TIA.
I really appreciated your review and evaluation of these products. Thanks
My pleasure!
Very useful knowledge. Please do another one in 6 - 12 months to see if any of them have improved.
That's a great idea
A truly brilliant comparison with good balanced scoring making it very accurate, judgement wise. The end of the video rightly calls out the continuing weakness of Firefly in this area - especially for any professional work. I look forward to more 'nailed it' comparisons. Succint and to the point. Brilliant.
Thanks! I really appreciate it. This was fun and I'm glad it's well received.
Great comparison, Blake. If only Magnific had a subscription level below Pro ($39/month). There's no way I'd do 200 upscales or generations a month (what's included in the Pro level). Wish they had a Basic level for more like $15-20/month.
I agree! I got it back in Dec 2023 and they gave me 30,000 credits. I am currently at 24,000. Many things require anywhere from 7 - 100 credits per generation depending on the resolution. So it is VERY difficult to hit that pro level even for someone like me who generates a lot of images.
@ Yikes! That's a lot of remaining credits at the end of the year. When it renews in a couple months you'll have about 53,000 credits. 😬 Too bad I can't buy $40 worth of credits and replenish them once I've used them up in 9 months. Wanna sell me some of your extra credits on the black market? Haha.
@bradphoto ha! Wish I could. I didn't know if they add up or reset. I'll find out in a month or so 🤣
They are owned by Freepik, it slow it is merging its features to Freepik. Soon will be just Freepik
A very well thought out and well presented look at these things. I know a-lot more now than I did 30 min ago. I'm so disappointed in Adobe with what you showed. And your comments around the 17 min mark are spot on. I use or at least try to use Gen remove in both Lr and PS often. Even after adobe max updates it's quite limited. Thanks - Tim
I'm glad you enjoyed it. It was a lot of fun rendering everything and putting this one together today. I appreciate your thoughts and open mind.
Enjoyable. So many AI generators to work with. Sometimes the errors or the misunderstanding can be the funniest things. I find that knowing PS and other Post processing software help tremedousely with the final results. Also MJ's new web based editor has some cool features. Thanks for putting this together was a fun and informative watch. :)
I agree. Knowing Ps makes AI image generation a much more enjoyable experience and you can edit them to your ideal vision. I haven't played with the web module for MJ, only the Discord.
Thanks Blake. I've just begun to play with these AI platforms. Your critique is very useful and on point.
Great review Blake. Thank you.
You're welcome! It was fun to make.
Thanks for the comparisons, have you any experience with these using your own photo then using AI to fill in the background with prompts like generative fill does in photoshop? I have noticed if I use prompts to give me an illustrative background it can work magic however for work i need a realistic background. Thanks for any suggestions.
I haven't used it in my work in that way. However, magnific would be best for that. Even if slightly illustrative, blurred a little it may look natural.
For fun I tried same wizard prompt in Midjourney and got pretty decent result - point is that I'm using my specific personalized styles. Without personalization styles images are ... OK but not that good.
Cool! Glad your styles are working 😁
Excellent presentation 😃
Thanks 😁 it was fun to put together.
One of the most enjoyable videos I've watched today Blake! Great comparison and I concur with your final decision.
Glad you enjoyed it. That Magnific is just incredible. I've been rendering stuff all day.
@ I’m getting closer and closer to getting it!
Great review Blake! Thanks. I mostly use Midjourney and have found if I say Photo or Photograph rather than Photo Realistic I get better results. Realistic seems to be some kind of trigger that it’s not a real photo but photo like.
Thanks! I'll try that one.
How many image can be generated with PRO plan... i can't find the answer anywhere?
You get credits. 30,000 when you become pro. It's 7 credits for 2k resolution images and 22 credits for 4k resolution images. Up sizing can be anywhere from 5 - 75 credits based on how large you upscale it.
I got my account December 2023 and still have over 20,000 credits. It's safe to say you can generate a LOT with the Pro account.
@ each. Month you got 30k or?
@haralambiew 30,000 for the entire year.
Sunrise AND sunset in the same prompt!!? What exactly are you measuring, how to create something from an internally contradictory description?
That was a typo, my bad, but hey it put the limits of the AI to the test and it still produced wonderful results... Shows that it's smart even when human input is NOT, lol
I know I’m off topic, but I love your shirt. 😎
It's a decent shirt, but you should meet the people that have it to me 😁
Just so you know, photorealistic is an art style which makes image closer to realism. If you want a realistic image, dont use photorealistic in the prompt.
Hmm, I'll give it a try. I use that in my prompts all the time. I find that if I dont I get an illustrative look. What's the alternative?
@@f64Academy just add realistic or cinematic or cinematic still etc. using words like portrait photography, nature photography etc. in the prompt will also help moving ai towards realistic images. Not every time though sometimes ai more bends towards art styles as per its understanding of the prompt.
@@raghuram2018 thanks, I appreciate the help.
Leonardo AI, is starting to look reasonable as well
There are so many out there, it's really up to what works best for the user and what matches their ideal style.
Nice work thru AI stuff, Blake. You are totally right. It will only get better, and it is here to stay. I think for compositing it is very usefull. What that can evolve to in 5 years is exiting to think of, it is only a coupple of years sincewe even heard about it.
Right? The future is exciting in this space! Especially when Adobe gets the memo 🤣
Where is the water in the waterfalls coming from in the first one?
Snow melt from the glacier to the left that is home to unicorns, dragons, and very large people.
@@f64Academy ahhh, my tunnel vision didn't see peripherally- so i missed all that lol
@Nazareth434 I almost did too, you have to look really hard.
For some reason I always get a slight variation of the same woman's face every time I use Magnific (Mystic/Flux). And I don't seem to be the only one who has this issue because I see her face everywhere in videos about Flux. I agree that the overall quality surpasses even Midjourney but unless I can figure out how to avoid seeing that same face over and over, I have to stick with Midjourney and then upscale in Magnific.
That's still the best plan, I think. there's more versatility in the prompts from MJ. So picking the one you want is a good option, then upscaling it with Magnific as MJ's upscale isn't the best.
@@f64Academy Now if Photoshop could just get a Midjourney or Flux plugin that bypasses Firefly, we'd be all set. I agree with everything you said about Firefly. Adobe needs to rethink things.
You nailed it.
hahahaha
@@f64Academy One last thing - I use all of these tools as well - and even a few others - on a near, daily basis. The one dark horse no ones talking about is "Visual Electric". I've been using it for about a Month now - and in MANY cases - it outperforms Midjourney or at the very least adheres to the prompt way better. At certain creative and/or cinematic images MJ still outshines the rest, but VE is my go to for people and my own stock-type images. Also, the tech help is super responsive and i LOVE the whole "whiteboard canvas" setup they have. Let's you explore lots of options for a given client all in one "page". (and NO, i don't work for them in any capacity and i don't get anything if you go check it out!)
Can I buy the "You nailed it" sticker, to put on my tripod next to the f64 one?
I'm already working on that 🤣
f64-academy.creator-spring.com/listing/nailed-it-die-cut-sticker lol
Great review but Adobe may getting a bad rap. I just downloaded Photoshop (beta) 25.1.0 and entered your prompts for each example in Generative Workspace. All four options from Photoshop's Firefly Image 3 were as good or better than the competitor's except for the "Nailed It" logo. Photoshop did a great job on the hammer and nail but messed up royally on the words "Nailed It" in each option.
Maybe in the direction of accuracy to prompt, but the details are always horrible and always will be until they release the 1024 restriction. It's not useable. I have the exact same version of PS installed and those were the exact images I received for my prompts and I made the canvas size 1024x768... All horrible.
Elon's Grok in X is excellent too.
I don't use X so I haven't seen that one, but thanks for letting me know.
@@f64Academy Actually, I take it back. Just used Magnific and it blew me away! Thanks.
Well this was interesting. I am blown away with how bad Adobe Firefly is. I use the generative fill to mostly do removal related tasks and maybe adding a plant here and there. Works okay for that but not great. With all the money adobe has coming in through subscriptions and all the photos they have access to via there not so user friendly EULA, and all the stock images they have access to on top of that, it is really pathetic what they have to offer. Looking forward to working with these AI tools over the winter.
I agree 100% it's crazy how bad it is for a company with seemingly unlimited resources.
5 for diversity on some of those photos lol 👏
It does a whole lot better at it than most humans, that's for sure.
Really good and it will improve.
Everyone else is, Adobe is behind the power curve on this one.
come on Adobe, it's the first noun in the promt and you whiffed 😂 you could have made a bouncy house castle and it would have been closer.
Haha! Right?!?
Magnific is expensive
Cost is relative. It's a lot cheaper than hiring someone or paying for stock.
Yes, it's true, these algorithms accidentally named AI will remain with us and will be increasingly perfect in creating imaginary worlds. However, nothing can replace a human in creativity, which is the food for these algorithms and there is no point in complaining because we are the ones who feed these algorithms with our creativity. Our activity has an impact on the quality of the images created. Complaining about AI, we complain about our own creativity, which people forget about. I do not use such options, but I follow its progress. Greetings from Poland from the city of "Przemyśl Fortress".
That's an interesting way to look at it. Thanks.
Adobe Firefly sucks. Pardon my French! It's helpful to see the differences.
It's literally the worst model out there.
If the text is off, it should get a ZERO.
Nah, I can edit that in PS. All AI images need some help in PS and that's an easy one to fix with the remove tool.
I tried the same comments in Photoshop 2025 and got way better results than you showed
That's good! I wish I did 🤣
@@f64Academy I was not intending to be a smart a... but the results were at least as good as chatgpt